The aim was to identify demographic and environmental factors that affect the incidence and course of systemic lupus erythematosus (SLE) by studying geographical variation in mortality from the disease. Counts of SLE deaths during a 10-y period were determined for each of the 3111 counties in the United States. The counts were subjected to statistical techniques for detecting spatial clusters with significantly higher or lower mortality than expected. Census data were used to compare attributes of the resulting clusters. Four clusters with significantly elevated SLE mortality were identified. They were centered in Alabama, Arkansas, Louisiana and New Mexico and had 41-65% more SLE deaths than expected. Four other clusters with significantly low mortality were centered in Minnesota, Vermont, Virginia and Washington and had 32-44% fewer deaths than expected. Comparison of SLE mortality rates between the New Mexico and Vermont clusters yielded relative risks of 3.2, 3.4 and 5.9 for white, black and other race women, respectively, and of 2.1, 5.2 and 'undefined' for white, black and other race men. Census data showed that the clusters with elevated mortality had higher poverty rates and/or greater concentrations of ethnic Hispanics than those with lower mortality. SLE mortality rates show great regional variation within the United States. The variation appears to reflect, in part, geographical patterns in socioeconomic status and in the distribution of persons of Hispanic origin.